CN112001433A - Flight path association method, system, equipment and readable storage medium - Google Patents
Flight path association method, system, equipment and readable storage medium Download PDFInfo
- Publication number
- CN112001433A CN112001433A CN202010807867.3A CN202010807867A CN112001433A CN 112001433 A CN112001433 A CN 112001433A CN 202010807867 A CN202010807867 A CN 202010807867A CN 112001433 A CN112001433 A CN 112001433A
- Authority
- CN
- China
- Prior art keywords
- data
- training
- cnn
- network
- track
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 68
- 238000012549 training Methods 0.000 claims abstract description 68
- 230000015654 memory Effects 0.000 claims abstract description 32
- 238000004422 calculation algorithm Methods 0.000 claims description 28
- 238000012360 testing method Methods 0.000 claims description 27
- 238000004590 computer program Methods 0.000 claims description 16
- 238000007781 pre-processing Methods 0.000 claims description 11
- 210000002569 neuron Anatomy 0.000 claims description 10
- 238000002372 labelling Methods 0.000 claims description 6
- 230000035945 sensitivity Effects 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 5
- 230000007787 long-term memory Effects 0.000 claims description 4
- 230000002441 reversible effect Effects 0.000 claims description 4
- 230000006403 short-term memory Effects 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 claims description 2
- 238000000605 extraction Methods 0.000 abstract description 5
- 230000007547 defect Effects 0.000 abstract description 2
- 238000012512 characterization method Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 7
- 238000004088 simulation Methods 0.000 description 6
- 238000011176 pooling Methods 0.000 description 5
- 239000000284 extract Substances 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000012937 correction Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000000342 Monte Carlo simulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a flight path association method, a system, equipment and a readable storage medium, which preprocess flight path data of different targets and the same motion mode; respectively constructing a convolutional neural network model and a long-time memory unit based on the preprocessed flight path data to form a CNN-LSTM network connected in parallel; training the CNN-LSTM network, firstly training the CNN network, and then training the LSTM based on the trained CNN network to obtain a trained CNN-LSTM network; verifying the trained CNN-LSTM network, and acquiring a track correlation result based on the verified CNN-LSTM network and the preprocessed track data; the method realizes the intelligent extraction of the spatial features and the time sequence features of the flight path data, avoids the defects of unobvious feature characterization or insufficient feature quantity caused by manually selecting the features, improves the accuracy of the flight path association, and provides an effective implementation method for solving the problem of the flight path association.
Description
Technical Field
The invention belongs to the technical field of communication, and relates to a track association method, a system, equipment and a readable storage medium, in particular to a track association method of a multi-target multi-sensor based on a convolutional neural network and a long-time memory unit network in deep learning, equipment based on the association method and a readable storage medium.
Background
With the continuous exploration of objective objects by people, due to the diversification of information expression forms, the large amount of information, the complexity of information relation and the timeliness of information processing, which completely exceed the capability of manually processing information, the information collected by a single sensor cannot meet the complete cognitive requirement of people on the objects, so that a plurality of sensors are required to collect a large amount of data information, and the information resources of the plurality of sensors are effectively fused to obtain the overall information of the target to the greatest extent. The information fusion principle is that the information content of a plurality of sensors is utilized, redundant information of the information in space and time is reasonably removed by defining a certain criterion, useful information is interactively fused, the aim is to separate observation information by using each sensor, and more effective information is derived by optimizing and combining the information. The information fusion technology is widely applied to civil fields and military fields such as finance, medical treatment, traffic control, tactical command control (TCAC) and the like.
The track association means whether the track information obtained by two or more sensors corresponds to the same target. The association process is a process of judging whether the information received by the sensor platform and the maneuvering target have the relationship, and information quantity and information relationship in data in the sensors are discovered as much as possible, so that the data can be better associated and processed. How to discover hidden relations among data in a large amount of track data collected by multiple sensors so as to reflect the target state more quickly, accurately and comprehensively is a key research. The existing traditional track association algorithm mainly describes the similarity of tracks based on a statistical method and a membership function in fuzzy mathematics, along with the development of science and technology and the optimization of sensor performance, the quantity of track data which can be acquired is gradually increased, and the quantity of information contained in the data is increased, so that the requirement on data processing capacity is gradually increased. Because the number of the maneuvering targets and the number of the sensors are gradually increased, the probability of the intersection and bifurcation of the flight path is increased, how to find the hidden feature relationship in a large amount of flight path data collected from a plurality of sensor platforms, perform proper feature extraction, perform more accurate association judgment according to the extracted feature relationship, and reduce the error interference of the association result in target tracking and information fusion, the problem of flight path association in a multi-sensor target environment needs to be solved.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a multi-target multi-sensor track correlation method based on parallel CNN-LSTM, which respectively constructs a convolutional neural network model (CNN) and a long-time memory unit (LSTM) to form a parallel CNN-LSTM network, and improves the accuracy of track correlation based on the CNN-LSTM network.
In order to achieve the purpose, the invention adopts the technical scheme that: a track association method, comprising the steps of:
preprocessing track data of different targets and the same motion mode;
respectively constructing a convolutional neural network model and a long-time memory unit based on the preprocessed flight path data to form a CNN-LSTM network connected in parallel;
training the CNN-LSTM network, firstly training the CNN network, and then training the LSTM based on the trained CNN network to obtain a trained CNN-LSTM network;
and verifying the trained CNN-LSTM network, and extracting spatial features and temporal features of the preprocessed track data based on the verified CNN-LSTM network to obtain a track correlation result.
The track data preprocessing specifically comprises the following steps: and generating a track data set with a set maneuvering target number by using an interactive multi-model algorithm, and randomly selecting 70% of the acquired data as a training data set and 30% of the acquired data as a testing data set.
The method for training the CNN-LSTM network specifically comprises the following steps:
step 301, determining a termination condition and a maximum iteration number T of convolutional neural network model training;
step 302, dividing the preprocessed flight path data into a training set and a testing set, randomly selecting flight path data from the training data set, using the flight path data as an input training sample, labeling the selected data in the training data set to indicate a target to which the flight path data belongs, and training a neural network;
step 303, calculating the output of each layer of the convolutional neural network by adopting a forward propagation algorithm through input training samples input into the convolutional neural network;
304, correcting the weight and the offset value of each layer of nodes of the convolutional neural network by adopting a back propagation algorithm through outputting a training sample and the output of each layer of the convolutional neural network, and checking the performance of the current network model through the track association correct rate;
and 305, repeatedly executing the step 302 to the step 304 until the sensitivity of the output layer of the convolutional neural network meets the requirement of the termination condition of the convolutional neural network model training or the repetition frequency is T-1, and storing the structure of the modified convolutional neural network and the weight and the offset value of each layer of node to obtain the trained convolutional neural network model.
Training the LSTM based on the trained CNN network comprises the following steps:
step 306, initializing the trained CNN network;
step 307, dividing the preprocessed flight path data into a training set and a testing set, randomly selecting flight path data from the training data set, taking the flight path data as an input training sample, labeling the data selected from the training data set to indicate a target to which the data belongs, and training the long-time memory unit;
308, adopting a forward propagation algorithm to calculate the output value of the neuron in the long-time memory unit in a forward direction;
step 309, calculating error items of the neurons in a reverse mode by adopting a reverse propagation algorithm, correcting weights and bias values of the neurons in the long and short term memory unit according to the corresponding error items, and checking the performance of the current network model through the track association accuracy;
and 310, repeatedly executing the steps 307-309 until the sensitivity of the convolutional neural network output layer meets the requirement of the termination condition of the convolutional neural network model training or the repetition frequency is T-1, and obtaining the trained long-time memory unit network.
The verification of the trained CNN-LSTM network and the acquisition of the track correlation result comprise the following steps:
step 41, dividing the preprocessed data into a training set and a testing set, randomly selecting a time domain signal from the testing set, taking the time domain signal as an input testing sample, and simultaneously selecting a corresponding label to compare subsequent network output results;
step 42, detecting the correctness of the network model by inputting the test sample into the CNN by adopting a forward propagation algorithm;
step 43, detecting the correctness of the correlation of the obtained track data by comparing the test sample label with the final output of the CNN-LSTM network by adopting a back propagation algorithm;
and 44, repeating the steps 41 to 43 until all the preprocessed data in the test set are selected, and counting whether the identification of each signal is correct or not to obtain the final track association correct probability.
The track data includes position information, speed information, acceleration information, and category information to which the data belongs at each time.
A track correlation system comprises a data preprocessing module, a model construction module, a model training module and a model verification and result output module;
the data preprocessing module is used for preprocessing the track data of different targets in the same motion mode;
the model building module respectively builds a convolutional neural network model and a long-time memory unit based on the preprocessed flight path data to form a parallel CNN-LSTM network;
the model training module trains the CNN-LSTM network, firstly trains the CNN network therein, and then trains the LSTM based on the trained CNN network to obtain the trained CNN-LSTM network;
and the model verification and result output module verifies the trained CNN-LSTM network and acquires a track correlation result based on the verified CNN-LSTM network and the preprocessed track data.
A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the track association method of the invention when executing the computer program.
A computer-readable storage medium, in which a computer program is stored, which, when being executed by a processor, carries out the steps of the track association method according to the invention.
Compared with the prior art, the invention has at least the following beneficial effects:
the method collects the track data information of the maneuvering target, trains and learns the constructed CNN and LSTM network models, extracts the characteristics of the track information on the space dimension and the time dimension on the basis of learning a large amount of data, extracts the characteristics of the track on the space dimension such as speed and acceleration on the CNN network, extracts the characteristics of the track on the LSTM network, extracts the characteristics of the track on the time dimension such as the time characteristic between each moment in a time sequence, avoids the loss of track data characteristic extraction caused by extracting the time characteristic after extracting the space characteristic of the track data information in the prior art, and improves the association accuracy of track association compared with the prior art; the invention directly uses the flight path data as the input of the CNN-LSTM network, and uses the corrected structure of the CNN-LSTM network and the weight and the offset value of each layer of nodes to extract the spatial characteristic and the time sequence characteristic of the flight path data, thereby avoiding the artificial analysis of the signal characteristic and the artificial determination of the selected signal characteristic in the prior art.
Drawings
FIG. 1 is a flow chart of a method that can be implemented in the present invention.
Fig. 2 shows a parallel CNN-LSTM network model structure for a specific application.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments.
Referring to fig. 1, a track association method based on parallel CNN-LSTM networks includes the following steps:
(1) preprocessing the flight path data:
acquiring a training data set and a testing data set:
and (3) generating a track data set with the maneuvering targets of 50 by using an interactive multi-model algorithm, and randomly selecting 70% of the acquired data as a training data set and 30% of the acquired data as a testing data set. In this embodiment, the number of sampling points of the flight path data is 60000;
(2) constructing a convolutional neural network model:
determining the number of input layer nodes, the number of output layer nodes, the number of convolutional layer, convolutional layer convolution kernel, pooling layer number and full-connection layer number of convolutional neural network, and the activation functions of convolutional layer, pooling layer and full-connection layer, and weighting value W of each layer nodelAnd offset blInitialization is performed.
In this embodiment, the total number of layers n of the convolutional neural network is 2, the number of nodes of the input layer is 60000, the number of nodes of the output layer is 50, the number of convolutional layers is 2, the number of pooling layers is 2, and the number of fully-connected layers is 1, as shown in fig. 2. The specific convolutional neural network model is constructed as follows:
net＝{i,c64,s,c128,s,f,o}
wherein i denotes that the layer is an input layer, c64Indicating that the layer is a convolutional layer with a convolutional kernel number of 64, c128This layer is denoted as convolutional layer, and the number of convolutional kernels is 128. s denotes that the layer is a pooled layer, f denotes that the layer is a fully connected layer, o denotes that the layer is an output layer, and the convolution kernel sizes of the 2 convolutional layers are 1 × 5 and 1 × 3, respectively. The regularization is performed by selecting L2, the downsampling size of the pooling layer is 1 multiplied by 2, dropout is started after each layer of full connection layer, the dropout rate is 0.5, and the learning rate in the network is set to be 0.01. Using [0,1]Weight W of each layer of nodes initialized by normal distributionlAnd offset blThe activation functions of the convolution layer, the pooling layer and the full-connection layer all adopt linear rectification functions, the expression is f (x) max (0, x), x is an independent variable, and the value of x is equal to the input value of the layer node where the activation function is located.
(3) Constructing a long-time memory unit network:
determining the number of input layer nodes, the number of output layer nodes and the number of long and short memory unit layers of the long and short memory unit network, and determining the weight W of each layer of nodeslAnd offset blInitialization is performed.
In this embodiment, the total number n of long and short term memory cells is 1 layer, the number of nodes in the input layer is 60000, the number of nodes in the output layer is 50, and [0,1 ] is used in the same manner as the convolutional neural network initialization]Weight W of normally distributed initialization neuron nodelAnd offset bl。
(4) Training the convolutional neural network model:
(4a) determining a termination condition and a maximum iteration number T of convolutional neural network model training;
(4b) randomly selecting flight path data from the training data set, using the flight path data as an input training sample, and labeling the data selected from the training data set to indicate a target to which the data belongs so as to train the neural network;
(4c) calculating the output of each layer of the convolutional neural network by adopting a forward propagation algorithm through input training samples input into the convolutional neural network;
(3d) correcting the weight and the offset value of each layer of nodes of the convolutional neural network by adopting a back propagation algorithm through outputting a training sample and the output of each layer of the convolutional neural network, and checking the performance of the current network model through the track correlation accuracy;
(4e) repeatedly executing the steps (4b) - (4d) until the sensitivity of the convolutional neural network output layer meets the requirement of the termination condition of the convolutional neural network model training or the repetition frequency is T-1, storing the modified structure of the convolutional neural network and the weight and the offset value of each layer of node to obtain a trained convolutional neural network model;
(5) training the long-time memory unit:
(5a) first, network initialization is performed.
(5b) Randomly selecting flight path data from the training data set, using the flight path data as an input training sample, and labeling the data selected from the training data set to indicate the target to train the long-time memory unit and the short-time memory unit;
(5c) and adopting forward propagation to calculate the output value of the neuron in the long-short time memory unit in the forward direction.
(5d) Using back propagation, the error term sigma of the neuron is calculated backn，
Correcting the weight and the offset value of the neuron in the long and short term memory unit according to the corresponding error item, and checking the performance of the current network model through the track correlation accuracy;
(5e) repeatedly executing the steps (5b) - (5d) until the sensitivity of the convolutional neural network output layer meets the requirement of the termination condition of the convolutional neural network model training or the repetition frequency is T-1, and obtaining a trained long-time memory unit network;
(6) and (3) carrying out performance test on the trained parallel CNN-LSTM network model:
(6a) randomly selecting a time domain signal from the test set, taking the time domain signal as an input test sample, and simultaneously selecting a corresponding label to compare the subsequent network output results;
(6b) detecting whether the establishment of the network model is correct or not by adopting a forward propagation algorithm and inputting an input test sample into the convolutional neural network;
(6c) comparing the test sample label with the final output of the CNN-LSTM network by adopting a back propagation algorithm to obtain whether the track data association is correct or not;
(6d) and (5) repeating the steps (6a) - (6c) until all the track data in the test set are selected, and counting whether the identification of each signal is correct or not to obtain the final track association correct probability.
The technical effect of the invention is further explained through simulation experiments.
1 simulation condition:
track data information of 50 maneuvering targets is adopted in simulation conditions, track information measurement is carried out on the maneuvering targets by 3 sensors, the total sampling point number of the track data is 60000, the size of a training set is 42000, the size of a testing set is 18000, and the number of Monte Carlo simulation tests is 100;
the convolutional neural network model is built in Keras of Python 3.6;
the number of training iterations is 50, the batch _ size is 64, the learning rate is 0.001, and the loss function is the coordinated _ cross function.
2, simulation content: the invention and the existing track association algorithm are adopted to carry out track association on 50 targets, and the association result is shown in table 1.
TABLE 1 comparison of results of different track association algorithms
| Track association algorithm | Weighting method | Correction method | K nearest neighbor domain method | Text algorithm | End-to-end algorithm |
| Correlation accuracy | 66.72％ | 68.86％ | 65.29％ | 91.67％ | 89.87％ |
3, simulation result analysis:
table 1 shows the correlation accuracy comparison result of the track correlation between the track correlation algorithm proposed herein and the existing track correlation algorithm. According to the simulation result, the parallel CNN-LSTM-based track association method has higher track association accuracy when the number of targets is 50 and the motion mode of the track data is the same. The weighting method, the correction method and the K-nearest neighbor domain method are all used for judging track association only through the spatial features of track data, the end-to-end algorithm is used for extracting the time sequence features after extracting the spatial features of the track data, and the time sequence features are incomplete due to the fact that partial feature vectors are lost in the feature extraction process of a convolutional neural network for extracting the spatial features, and subsequent feature extraction and analysis are affected.
The invention also provides a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the track association method of the invention when executing the computer program.
The inventive track association methods may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects, and may take the form of a computer program product embodied on one or more computer-usable storage media including, but not limited to, disk storage, CD-ROM, optical storage, and the like, having computer-usable program code embodied in the medium. The track association method of the present invention, if implemented in the form of software functional units and sold or used as a stand-alone product, can be stored in a computer-readable storage medium.
Based on such understanding, in the exemplary embodiment, a computer readable storage medium is also provided, all or part of the processes in the method of the above embodiments of the present invention can be realized by a computer program to instruct related hardware, the computer program can be stored in the computer readable storage medium, and when the computer program is executed by a processor, the steps of the above method embodiments can be realized. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. Computer-readable storage media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. It should be noted that the computer readable medium may contain content that is subject to appropriate increase or decrease as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media does not include electrical carrier signals and telecommunications signals as is required by legislation and patent practice. The computer storage medium may be any available medium or data storage device that can be accessed by a computer, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.), optical memory (e.g., CD, DVD, BD, HVD, etc.), and semiconductor memory (e.g., ROM, EPROM, EEPROM, nonvolatile memory (NANDFLASH), Solid State Disk (SSD)), etc.
In an exemplary embodiment, a computer device is also provided, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the track association method when executing the computer program. The processor may be a Central Processing Unit (CPU), other general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc.
The above-mentioned contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modification made on the basis of the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (9)
1. A flight path correlation method is characterized by comprising the following steps:
preprocessing track data of different targets and the same motion mode;
respectively constructing a convolutional neural network model and a long-time memory unit based on the preprocessed flight path data to form a CNN-LSTM network connected in parallel;
training the CNN-LSTM network, firstly training the CNN network, and then training the LSTM based on the trained CNN network to obtain a trained CNN-LSTM network;
and verifying the trained CNN-LSTM network, and extracting spatial features and temporal features of the preprocessed track data based on the verified CNN-LSTM network to obtain a track correlation result.
2. The track association method according to claim 1, wherein the track data preprocessing specifically comprises: and generating a track data set with a set maneuvering target number by using an interactive multi-model algorithm, and randomly selecting 70% of the acquired data as a training data set and 30% of the acquired data as a testing data set.
3. The track correlation method according to claim 1, wherein the training of the CNN-LSTM network specifically comprises the steps of:
step 301, determining a termination condition and a maximum iteration number T of convolutional neural network model training;
step 302, dividing the preprocessed flight path data into a training set and a testing set, randomly selecting flight path data from the training data set, using the flight path data as an input training sample, labeling the selected data in the training data set to indicate a target to which the flight path data belongs, and training a neural network;
step 303, calculating the output of each layer of the convolutional neural network by adopting a forward propagation algorithm through input training samples input into the convolutional neural network;
304, correcting the weight and the offset value of each layer of nodes of the convolutional neural network by adopting a back propagation algorithm through outputting a training sample and the output of each layer of the convolutional neural network, and checking the performance of the current network model through the track association correct rate;
and 305, repeatedly executing the step 302 to the step 304 until the sensitivity of the output layer of the convolutional neural network meets the requirement of the termination condition of the convolutional neural network model training or the repetition frequency is T-1, and storing the structure of the modified convolutional neural network and the weight and the offset value of each layer of node to obtain the trained convolutional neural network model.
4. The track correlation method of claim 1, wherein training the LSTM based on the trained CNN network comprises the steps of:
step 306, initializing the trained CNN network;
step 307, dividing the preprocessed flight path data into a training set and a testing set, randomly selecting flight path data from the training data set, taking the flight path data as an input training sample, labeling the data selected from the training data set to indicate a target to which the data belongs, and training the long-time memory unit;
308, adopting a forward propagation algorithm to calculate the output value of the neuron in the long-time memory unit in a forward direction;
step 309, calculating error items of the neurons in a reverse mode by adopting a reverse propagation algorithm, correcting weights and bias values of the neurons in the long and short term memory unit according to the corresponding error items, and checking the performance of the current network model through the track association accuracy;
and 310, repeatedly executing the steps 307-309 until the sensitivity of the convolutional neural network output layer meets the requirement of the termination condition of the convolutional neural network model training or the repetition frequency is T-1, and obtaining the trained long-time memory unit network.
5. The track association method as claimed in claim 1, wherein the step of verifying the trained CNN-LSTM network to obtain the track association result comprises the steps of:
step 41, dividing the preprocessed data into a training set and a testing set, randomly selecting a time domain signal from the testing set, taking the time domain signal as an input testing sample, and simultaneously selecting a corresponding label to compare subsequent network output results;
step 42, detecting the correctness of the network model by inputting the test sample into the CNN by adopting a forward propagation algorithm;
step 43, detecting the correctness of the correlation of the obtained track data by comparing the test sample label with the final output of the CNN-LSTM network by adopting a back propagation algorithm;
and 44, repeating the steps 41 to 43 until all the preprocessed data in the test set are selected, and counting whether the identification of each signal is correct or not to obtain the final track association correct probability.
6. The track correlation method according to claim 1, wherein the track data includes position information, velocity information, acceleration information, and category information to which the data belongs at each time.
7. A track correlation system is characterized by comprising a data preprocessing module, a model building module, a model training module and a model verification and result output module;
the data preprocessing module is used for preprocessing the track data of different targets in the same motion mode;
the model building module respectively builds a convolutional neural network model and a long-time memory unit based on the preprocessed flight path data to form a parallel CNN-LSTM network;
the model training module trains the CNN-LSTM network, firstly trains the CNN network therein, and then trains the LSTM based on the trained CNN network to obtain the trained CNN-LSTM network;
and the model verification and result output module verifies the trained CNN-LSTM network and acquires a track correlation result based on the verified CNN-LSTM network and the preprocessed track data.
8. A computer arrangement comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the track association method as claimed in any one of claims 1 to 6 when executing the computer program.
9. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the track association method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010807867.3A CN112001433A (en) | 2020-08-12 | 2020-08-12 | Flight path association method, system, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010807867.3A CN112001433A (en) | 2020-08-12 | 2020-08-12 | Flight path association method, system, equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112001433A true CN112001433A (en) | 2020-11-27 |
Family
ID=73462495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010807867.3A Pending CN112001433A (en) | 2020-08-12 | 2020-08-12 | Flight path association method, system, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112001433A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113068131A (en) * | 2021-06-01 | 2021-07-02 | 浙江非线数联科技股份有限公司 | Method, device, equipment and storage medium for predicting user movement mode and track |
| CN113239775A (en) * | 2021-05-09 | 2021-08-10 | 西北工业大学 | Method for detecting and extracting flight path in azimuth history map based on layered attention depth convolution neural network |
| CN113536655A (en) * | 2021-04-07 | 2021-10-22 | 北京聚树核科技有限公司 | Artificial intelligent deviation rectifying method and device for heliostat, electronic equipment and storage medium |
| CN115512221A (en) * | 2022-09-22 | 2022-12-23 | 中国人民解放军海军航空大学 | GNN-based synchronous track robustness correlation method |
| CN116451177A (en) * | 2023-06-15 | 2023-07-18 | 创意信息技术股份有限公司 | Track association method and device |
| CN116663863A (en) * | 2023-07-28 | 2023-08-29 | 石家庄科林电气股份有限公司 | Virtual power plant load prediction method based on scheduling plan |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170103309A1 (en) * | 2015-10-08 | 2017-04-13 | International Business Machines Corporation | Acceleration of convolutional neural network training using stochastic perforation |
| CN107463967A (en) * | 2017-08-25 | 2017-12-12 | 中国人民解放军海军航空工程学院 | Multi-source track association machine learning system |
| CN109633589A (en) * | 2019-01-08 | 2019-04-16 | 沈阳理工大学 | The Multi-target Data Associations assumed are optimized based on multi-model more in target following |
| CN110321401A (en) * | 2019-07-09 | 2019-10-11 | 中国人民解放军海军航空大学 | Space-time data is associated with deep learning method |
| CN110427893A (en) * | 2019-08-06 | 2019-11-08 | 西安电子科技大学 | A kind of specific emitter identification method, apparatus and computer storage medium based on convolutional neural networks |
-
2020
- 2020-08-12 CN CN202010807867.3A patent/CN112001433A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170103309A1 (en) * | 2015-10-08 | 2017-04-13 | International Business Machines Corporation | Acceleration of convolutional neural network training using stochastic perforation |
| CN107463967A (en) * | 2017-08-25 | 2017-12-12 | 中国人民解放军海军航空工程学院 | Multi-source track association machine learning system |
| CN109633589A (en) * | 2019-01-08 | 2019-04-16 | 沈阳理工大学 | The Multi-target Data Associations assumed are optimized based on multi-model more in target following |
| CN110321401A (en) * | 2019-07-09 | 2019-10-11 | 中国人民解放军海军航空大学 | Space-time data is associated with deep learning method |
| CN110427893A (en) * | 2019-08-06 | 2019-11-08 | 西安电子科技大学 | A kind of specific emitter identification method, apparatus and computer storage medium based on convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 翁建新等: "利用并联CNN-LSTM的调制样式识别算法", 《信号处理》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113536655A (en) * | 2021-04-07 | 2021-10-22 | 北京聚树核科技有限公司 | Artificial intelligent deviation rectifying method and device for heliostat, electronic equipment and storage medium |
| CN113536655B (en) * | 2021-04-07 | 2022-07-22 | 北京聚树核科技有限公司 | Heliostat artificial intelligence deviation rectifying method, heliostat artificial intelligence deviation rectifying device, electronic equipment and storage medium |
| CN113239775A (en) * | 2021-05-09 | 2021-08-10 | 西北工业大学 | Method for detecting and extracting flight path in azimuth history map based on layered attention depth convolution neural network |
| CN113068131A (en) * | 2021-06-01 | 2021-07-02 | 浙江非线数联科技股份有限公司 | Method, device, equipment and storage medium for predicting user movement mode and track |
| CN115512221A (en) * | 2022-09-22 | 2022-12-23 | 中国人民解放军海军航空大学 | GNN-based synchronous track robustness correlation method |
| CN115512221B (en) * | 2022-09-22 | 2024-02-27 | 中国人民解放军海军航空大学 | GNN-based synchronous track robustness association method |
| CN116451177A (en) * | 2023-06-15 | 2023-07-18 | 创意信息技术股份有限公司 | Track association method and device |
| CN116451177B (en) * | 2023-06-15 | 2023-09-12 | 创意信息技术股份有限公司 | Track association method and device |
| CN116663863A (en) * | 2023-07-28 | 2023-08-29 | 石家庄科林电气股份有限公司 | Virtual power plant load prediction method based on scheduling plan |
| CN116663863B (en) * | 2023-07-28 | 2023-10-20 | 石家庄科林电气股份有限公司 | Virtual power plant load prediction method based on scheduling plan |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112001433A (en) | Flight path association method, system, equipment and readable storage medium | |
| CN113378632B (en) | Pseudo-label optimization-based unsupervised domain adaptive pedestrian re-identification method | |
| CN108596327B (en) | Seismic velocity spectrum artificial intelligence picking method based on deep learning | |
| CN111913156A (en) | Radar radiation source individual identification method based on deep learning model and feature combination | |
| CN108447057B (en) | SAR image change detection method based on significance and depth convolution network | |
| CN110020712B (en) | Optimized particle swarm BP network prediction method and system based on clustering | |
| CN111368930B (en) | Radar human body posture identification method and system based on multi-class spectrogram fusion and hierarchical learning | |
| CN112949383A (en) | Waveform agility radar radiation source identification method based on Hydeep-Att network | |
| CN112949391A (en) | Intelligent security inspection method based on deep learning harmonic signal analysis | |
| CN114675249A (en) | Attention mechanism-based radar signal modulation mode identification method | |
| KR102407835B1 (en) | Method and apparatus for classifying pulse radar signal properties based on machine learning | |
| CN114973019A (en) | Deep learning-based geospatial information change detection classification method and system | |
| CN114594440A (en) | Radar high-resolution one-dimensional range profile target identification method and system based on double parallel networks | |
| CN113051983A (en) | Method for training field crop disease recognition model and field crop disease recognition | |
| CN112859034B (en) | Natural environment radar echo amplitude model classification method and device | |
| CN110969203B (en) | HRRP data redundancy removing method based on self-correlation and CAM network | |
| CN113239809A (en) | Underwater sound target identification method based on multi-scale sparse SRU classification model | |
| CN117110446A (en) | Method for identifying axle fatigue crack acoustic emission signal | |
| CN116415505A (en) | System fault diagnosis and state prediction method based on SBR-DBN model | |
| KR102407834B1 (en) | Method and apparatus for property-based classification of long-pulse radar signals | |
| Woolford et al. | Convergent data sharpening for the identification and tracking of spatial temporal centers of lightning activity | |
| Notkin et al. | Classification of Ground Moving Radar Targets with RBF Neural Networks. | |
| RU2819570C1 (en) | Method and device for recognizing type of radar station by its radiation | |
| CN114638298B (en) | Aircraft attack behavior prediction method and electronic equipment | |
| Wei et al. | Accretionary Learning With Deep Neural Networks With Applications |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201127 |
|
| RJ01 | Rejection of invention patent application after publication |